MySQL Fabric Client Cache Problem

Every now and again problems crop up and what's more is necessary to write to the Mysql support.

Following you'll find the description of the problem and the request to the Mysql Support.

I'll let you know the solution.

See you soon

Byeee


Hi,
I've been working on the MySQL Fabric framework lately and what's more I've made a cluster up. 


The cluster was made up of a Master (Node1), a Slave (Node2) and what's more a Fabric node. 


While I was looking at various failover scenarion a problem cropped up. 


Let me explain what the problem was. 


STEP 1 : The master (Node1) and Slave (node2) are online. The Jmeter client asks a connection to the Fabric node through the port 32274 and it commits a row on the Master which is replicated on the Slave afterwards. 


STEP 2: The Master (Node1) falls down, the Fabric node promotes the Slave (Node2) as a Master node and what's more marks the Node1 as a faulty node.
Be careful : The Jmeter client is still online. 


STEP 3: The Node1 , the previous Master, reboots and comes online again. Since the Fabric node has marked it as a faulty node it is still out of the cluster which is made up of one node (Node2) only. 


STEP 4: Again the Jmeter client execute an insert with the same connection string ( "jdbc:mysql:fabric://mysql.fabric.node:32274/DB?fabricUsername=admin&fabricPassword=xxxxx&fabricServerGroup=group_mob_pre-prod-2&fabricReportErrors=true") and that command will end up inserting a new row onto the faulty node (Node1). 


RESULT: I lost a row and what's more the Master (Node2) wasn't sincronyzed anymore. 


Do you happen to know if there is a way to work around that problem ? 


I've already tried the option [TTL=0] of the Fabric node but it doesn't work. 


If you get back on this to me, I'll appreciate it. 


Cheers